Portfolio failure

Financial Independence/Retire Early -- Learn How!
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Choosing an SWR and withdrawal strategy for 40 years is a bit like choosing a compatible wife for the next 40 years

KenM:

The point you are making here is that it is possible that a SWR analysis may not work. That is understood. That is why the caveat says "if the future is like the past." If the future is not like the past, all bets are off.

A concern that the future will not be like the past is not a justification for performing an invalid analysis of what the safe withdrawal rate is. In the event that the future is like the past, the valuation level that applies at the time of retirement will affect the investment result achieved. So that factor should be included in a valid analysis of what it is safe to do.

By no means am I saying that everyone should use the number produced by a valid SWR analysis as his or her personal withdrawal rate. If you are pessimistic about the future, you might want to make your number the SWR minus one percent. If you are optimistic, you might want to make your number the SWR plus one percent. It is an entirely reasonable procedure to elect a different withdrawal percentage than what the data indicates in deference to a personal belief that a data-based approach does not provide all the answers.

It is not proper, however, to include such considerations in the performance of the analysiis generating the data-based number. Adding subjective factors at that stage undermines the integrity of the analytic process. The right way to do this is to set up the analysis so that it reveals what the data actually says and then make whatever adjustments you want to make in accord with your personal subjective inclinations.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

hocus
The right way to do this is to set up the analysis so that it reveals what the data actually says and then make whatever adjustments you want to make in accord with your personal subjective inclinations.

This is critically important and it is very hard to do.

It is a discipline that you have to force yourself to do. It becomes easier over time as you see the benefits of doing things right.

My best examples from personal experience have to do with choosing which contractors to work on highly complex, new developments. Everything is done in a goldfish bowl environment. In several instances I have been surprised at the final answers. But every time I knew exactly why they were the best choices and that they would stand up to scrutiny.

Have fun.

John R.
raddr
*** Veteran
Posts: 265
Joined: Mon Nov 25, 2002 3:25 am
Contact:

Post by raddr »

hocus wrote:
It is not proper, however, to include such considerations in the performance of the analysiis generating the data-based number. Adding subjective factors at that stage undermines the integrity of the analytic process. The right way to do this is to set up the analysis so that it reveals what the data actually says and then make whatever adjustments you want to make in accord with your personal subjective inclinations.


So if you distrust the past how do you create an data-based SWR number for the future? Seems to me that you would by definition be generating a subjective number. You seem relatively certain that a "data-based" number can be generated. What is your formula?
User avatar
BenSolar
*** Veteran
Posts: 242
Joined: Mon Nov 25, 2002 5:46 am
Location: Western NC

Post by BenSolar »

hocus wrote: It is not proper, however, to include such considerations in the performance of the analysiis generating the data-based number. Adding subjective factors at that stage undermines the integrity of the analytic process. The right way to do this is to set up the analysis so that it reveals what the data actually says and then make whatever adjustments you want to make in accord with your personal subjective inclinations.


But the very act of selecting 'data' on the new parameters we want to add to the study is a subjective act, because, the data doesn't really exist. We don't have data on the SWR from PE-10 of low 40s. A simple extrapolation of a straight regression line would make the SWR strongly negative for a 75/25 S&P/cash port starting from those valuations, which is clearly nonsensical unless you are using leverage. We don't have extended return histories on REITs, small cap value, local real estate, etc., to plug into our simulations.

So we make assumptions, which are necessarily subjective.

I know you know this, hocus, but you've been hammering so hard on this 'objective calculation' based on the future being like the past, that I think some people are just thinking 'impossible, to many assumptions.' I think that JWR's definition is a good base for mutual understanding.

What we get out of advanced calculations of SWR like raddr does is an improvement (we hope ) on the Trinity style studies, but it will still be used as a rule of thumb. :)

Ben
"Do not spoil what you have by desiring what you have not; remember that what you now have was once among the things only hoped for." - Epicurus
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

You seem relatively certain that a "data-based" number can be generated. What is your formula?

We need to distinguish between things we know for certain and things we do not know for certain. If we were aware of every single factor that affects the SWR and we possessed perfect data showing the precise effect of each of those factors, this would be a piece of pie. We are not in that happy circumstance. We have to make do with the level of knowledge of what affects SWRs that is available to us today and with the quality of data available to us today. We have no choice in the matter..

However, the fact that as a practical matter we cannot design a perfect SWR analysis today does not justify deliberately designing the tool in ways that make it less effective than it could otherwise be. We know for a fact that volatility affects the result. Thus, a methodology that makes no attempt to incorporate the effects of volatility is invalid, in my view. We also know that changes in valuation levels affect the result. So, as with volatility, there is no excuse for ignoring this effect. A study methodology that ignores it, now that we know it always has an effect, can fairly be termed "invalid."

We do not yet have perfect knowledge of how to measure the effect of changes in valuation levels. Bernstein expressed great confidence in the long-term accuracy of the Gordon Equation. If you believe what he says in Chapter Two of "The Four Pillars of Investing" you will vote for using the Gordon Equation to assess the effect of changes in valuation levels. It is my understanding that JWR1945 is not as keen as Bernstein re the Gordon Equation, and it appears that Gummy is not as keen either. I respect the views of Bernstein, JWR1945, and Gummy, and I am no investment expert myself. So all that I can say on this question of the Gordon Equation is that it is worth looking at it in more depth. I do not feel confident enough one way or the other saying today that we should use the Gordon Equation or not use it.

If we find that some approach to assessing the effects of changes in valuation other than the Gordon Equation makes more analytical sense, we should go with some alternative. JWR1945 used an alternative approach in his independent assessment of the SWR for the year 2000. I proposed yet another alternative in my May 13, 2002, post. My initial thought was that we would calculate three SWRs: (1) the SWR that applies at times of high valuation; (2) the SWR that applies at times of medium valuation; and (3) the SWR that applies at times of how valuation. There are probably other reasonable approaches that other reasonable people could come up with.

The one approach that is not reasonable is to continue to employ the conventional methodology, a methodology that makes no attempt to assess the effects of changes in valuation levels whatsoever. That we know is wrong. An approach that does not even aim to incorporate all the effects that bear on the question is invalid, in my view.

We haven't examined the question of how to incorporate the effects of changes in valuation levels in enough depth to make a decision today. There is room for reasonable differences of opinion on this "how" question.
What I am saying is that there is not room for reasonable differences of opinion on the "Whether to" question. We know as a matter of mathematical certainty that changes in valuation levels affect SWRs. So it is just wrong not to account for this factor.

Will the number we come up with be correct in a metaphysical sense? Perhaps not. But it sure as shootin' will be more correct than any number we would come up with using the conventional methodology. That methodology does not factor in the effects of changes in valuation levels at all. We know for certain that the effect is not zero. So why perform the calculation using a methodology that contains an unspoken presumption that it is?
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

A simple extrapolation of a straight regression line would make the SWR strongly negative for a 75/25 S&P/cash port starting from those valuations, which is clearly nonsensical unless you are using leverage.

I do not believe that any of what you are pointing to here comes into play if you use the Gordon Equation to make the adjustment. I believe that it is only a PE-based adjustment that would give rise to the sorts of complications you are pointing to. It doesn't appear to me that the complications are insurmountable in any event.

We don't have extended return histories on REITs, small cap value, local real estate, etc., to plug into our simulations.

Each asset class is different. Different sorts of data are needed to support findings in regard to different sorts of asset classes. For example, we do not have much experience with TIPS. But the characteristics of the asset class make it particularly easy to assess the results likely to be achieved from investment in that investment class. We need to take each asset class one by one, and figure out what needs to be done to analyze it, or, alternatively, reach a decision that it is not possible to analyze it in a way providing reasonable assessments of the SWR for that class.

You can only do the best that you can do. I see no possibility of us doing more than it is as a practical matter feasible for us to do. I see no justification for us doing less than what as a practical matter it is feasible for us to do.

So we make assumptions, which are necessarily subjective.

You are correct that there are subjective assumptions that must be made as a practical matter. I do not deny this. But subjective assumptions are not something that we should seek out in circumstances in which they can be avoided. Subjectivity is a practical reality of SWR analysis in the real world, but it is not a good thing at the calculation stage of the process.

The argument that I have been making is that, when you know with certainty that a particular factor affects the question being posed, there is no valid analytical reason for excluding consideration of that factor from the analysis. To do that adds unnecessary subjectivity. It causes a deliberate diminishment of the accuracy of the number produced.

We should be aiming for as little subjectivity in the calculation as possible. The fact that we cannot get to zero subjectivity is not a rationale for permitting subjectivity in places where it can be easily avoided. We cannot avoid the subjectivity involved in choosing among the alternate ways of factoring in the valuation effect. But we should oppose methodologies that make subjective assumptions that are in direct conflict with things we know with mathematical certainty to be investment realities.

I know you know this, hocus, but you've been hammering so hard on this 'objective calculation' based on the future being like the past, that I think some people are just thinking 'impossible, to many assumptions.'

I understand the practical realities, of course. What I do not understand is the reluctance to including in the analysis a factor that always affects the result. A factor like that should always be given consideration.

I think that JWR's definition is a good base for mutual understanding.

I can't quote the statement at the moment, but I believe that the most recent JWR1945 statement described the inclusion of the valuation effect as a "necessary correction" to the conventional methodology, or something close to that. That statement I am OK with. It is fair to say that inclusion of this factor is a necessary correction to the conventional methodology.

JWR1945 is concerned that we not discard existing studies. I see no purpose served by discarding them. I agree with JWR1945 that it is possible to learn from them by examining the various ins and outs of them. My concern is that the numbers produced by the existing studies not be described to investors as "safe" numbers. It is reasonable to use the existing studies for analytical purposes, but not for investment purposes. They do not reveal the "safe" number for investment purposes.

There is nothing wrong with an individual investor using the 4 percent number produced by some of the existing studies. Deciding on what number to use in one's plan is a subjective exercise, so that does not concern me. My focus here is a deep-felt concern that it is a dangerously misleading statement to say that the historical data indicates that the 4 percent number is always safe. It is not a safe take-out number in many circumstances, and it is wrong for us to encourage in any way statements that it is. Quite to the contrary, we should be doing what we can to underline the most important thing that we have learned here--that the conventional methodology at times generates numbers that are not even close to the numbers generated by a methodology designed to reveal what the historical data indicates is safe.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

The one approach that is not reasonable is to continue to employ the conventional methodology, a methodology that makes no attempt to assess the effects of changes in valuation levels whatsoever. That we know is wrong. An approach that does not even aim to incorporate all the effects that bear on the question is invalid, in my view.

We haven't examined the question of how to incorporate the effects of changes in valuation levels in enough depth to make a decision today. There is room for reasonable differences of opinion on this "how" question.
What I am saying is that there is not room for reasonable differences of opinion on the "Whether to" question. We know as a matter of mathematical certainty that changes in valuation levels affect SWRs. So it is just wrong not to account for this factor.

Will the number we come up with be correct in a metaphysical sense? Perhaps not. But it sure as shootin' will be more correct than any number we would come up with using the conventional methodology. That methodology does not factor in the effects of changes in valuation levels at all. We know for certain that the effect is not zero. So why perform the calculation using a methodology that contains an unspoken presumption that it is?


someone suggested that if you are concerned about extreme valuations you could just adjust the rule of thumb on the basis of the degree of overvaluation. Seems like it has a lot of merit in terms of simplicity. It avoids the "math black box" but either approach seems to be based on assumptions.
Have fun.

Ataloss
User avatar
BenSolar
*** Veteran
Posts: 242
Joined: Mon Nov 25, 2002 5:46 am
Location: Western NC

Post by BenSolar »

hocus wrote: The argument that I have been making is that, when you know with certainty that a particular factor affects the question being posed, there is no valid analytical reason for excluding consideration of that factor from the analysis. To do that adds unnecessary subjectivity. It causes a deliberate diminishment of the accuracy of the number produced.

I think that no one here will argue this point.

Ben
"Do not spoil what you have by desiring what you have not; remember that what you now have was once among the things only hoped for." - Epicurus
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Either approach seems to be based on assumptions.

All methodologies are based on assumptions. The methodology that I am supporting is based on an assumption we know to be true--that changes in valuation levels affect the SWR. The conventional methodology is based on an assumption we know to be false--that changes in valuation levels do not affect the SWR. I think that it is better to go with an assumption that we know to be true than with an assumption that we know to be false.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

hocus: When you know with certainty that a particular factor affects the question being posed, there is no valid analytical reason for excluding consideration of that factor from the analysis.

BenSolar: I think that no one here will argue this point.

It's encouraging to hear you say that.
Post Reply